🆘 Plan B for your Claude memory

Don't let Anthropic
delete your AI memory.

Every conversation you've had with Claude — Code, Cowork, every project — sits on Anthropic's mercy. Random bans happen. claude-backup keeps a local copy on your disk and lets you hand the lot to ChatGPT, Cursor, or any other AI agent — in one command — when you need a Plan B.

View on GitHub

Free • MIT license • Works on macOS, Linux, Windows (WSL)

Install in 3 steps. No terminal needed.

You don't have to read any documentation. Paste the prompt into Claude (or ChatGPT, or Cursor) and the AI walks you through everything.

1
📋

Copy the install prompt

Click the big orange button above. The whole install instruction (a few hundred words) goes to your clipboard.

2
💬

Paste into any AI chat

Open Claude.ai, ChatGPT, Cursor, or your existing Claude Code window. Paste, hit send. The AI starts walking you through it.

3
✅

Follow what it tells you

The AI checks your Python, clones the repo, installs claude-backup, and runs your first backup. ~5 minutes total. You end up with all your conversations as readable Markdown files.

What it actually does

Four things, one tool. All offline, nothing sent anywhere.

💾

Local backup

Both Claude Code AND Claude Cowork sessions, exported to clean Markdown. Open in Obsidian, search with grep, archive in git.

🔀

Handoff to any AI

One command produces a paste-ready prompt for ChatGPT, Cursor, Claude.ai web — any chat agent. Continue your conversation there with full context.

🆘

Account-rescue bundle

Banned by Anthropic? Run claude-backup rescue. You get a portable folder with every session and a master meta-prompt for the new agent. Working again in 10 minutes.

🪄

Lazy install via AI

No terminal expertise required. The install prompt above turns any AI assistant into your installer. It checks Python, clones, installs, and runs your first backup with you.

The day Anthropic suspends your account, you have ~20 seconds to decide what to do.

Reddit and Hacker News are full of legitimate users who got auto-flagged: paying customers, people who used a VPN, people who paid through "wrong" channels. Appeals take weeks or months.

The thing they don't tell you: Anthropic only revokes API access. Your local files in ~/.claude/projects/ and Cowork's data folder are still on your disk. They're yours, not theirs.

One command — claude-backup rescue — packages everything into a portable bundle. You paste one file into ChatGPT (or any other AI) and it takes over from Claude with full context.

Don't wait until you're banned. It costs nothing to install today.

The actual commands (just so you know)

Honestly, you don't need to remember these. Paste the install prompt into Claude and it'll guide you. But here's what's under the hood.

$ claude-backup list # show every Claude Code & Cowork session $ claude-backup export-all -o ~/backups/ # back up everything to Markdown $ claude-backup handoff f7a07eec | pbcopy # copy a paste-ready prompt for another AI $ claude-backup rescue # 🆘 escape hatch when you're banned

FAQ

I'm not a developer. Can I really use this?
Yes — that's the whole point. Click the big orange button to copy the install prompt. Paste it into any AI chat (Claude.ai, ChatGPT, Cursor — whichever you already have open). The AI does all the work: checks your Python, runs the install, runs your first backup. You just answer questions when it asks. ~5 minutes.
Will Anthropic actually ban me?
Maybe, maybe not. But automated reviews flag legitimate users every week — paying customers, VPN users, people who paid through non-standard channels. Once the bot decides, you're out. Appeals take weeks. The whole point of this tool is: you don't have to find out the hard way that 3 months of Claude conversations were ONLY in their cloud.
Does it send my data anywhere?
No. Everything runs locally. The tool reads files that are already on your disk and writes Markdown files next to them. There's no telemetry, no analytics, no server contact. The source is on GitHub — go look.
What if I'm already banned?
Good news: your local files are still there. Anthropic only revokes API access; they can't reach into your disk. Install claude-backup, run claude-backup rescue, and you'll have a bundle you can paste into ChatGPT, Cursor, OpenClaw, or any other AI to keep working.
What does it cost?
Nothing. It's free, open source, MIT licensed. The author is parallelclaw — the source code is on GitHub. Use it, fork it, ship a better one.
✅ Copied. Now paste into any AI chat.